Labsco
MCP SERVER

Chess.com MCP (interactive views)

by nbialk ↗

Chess.com lookups that come back as a playable board — player ratings, a replayable last game, and the daily puzzle you can actually solve in the chat.

Games, Entertainment & Fandom
Summary
Worth opening for the pattern as much as for the chess.

Three tools is a small surface, and the point is what comes back: a puzzle you play against inside the conversation rather than a FEN string you are asked to picture. If you are building an MCP app with interactive views, this is a complete working example with the deployment path documented; if you just want chess, the hosted URL works with no setup at all.

What it is

A reference example of an MCP app where each tool ships its own interactive React view instead of returning plain text. It reads Chess.com, and the results render as a board you can step through — built with Skybridge, and running in ChatGPT and Claude.

What you get
  • A player profile from a username, with rapid, blitz and bullet ratings alongside the win, loss and draw record — `get-chess-player`
  • That player's most recent game — result, opponent and opening — with an interactive board replay rather than a move list — `get-last-game`
  • The Chess.com daily puzzle as a live board: solve it move by move with feedback as you go, reveal the solution, or reset and retry — `get-daily-puzzle`
Requirements

Nothing, if you use the hosted copy: add `https://mcp.chess.niklas.sh/mcp` to any client that supports remote MCP servers — no account, no key. To run it yourself you need Node.js 24.14.1+ and pnpm 9+; `pnpm install` then `pnpm dev` serves the MCP endpoint at `http://localhost:3000/mcp` with a DevTools UI on the same port. The package is `chess-mcp` (0.4.0 in package.json). Analytics are wired as middleware but stay a no-op unless `POSTHOG_API_KEY` is set, so a fork sends nothing.

Setup effort

One command — https://mcp.chess.niklas.sh/mcp